As machine learning methods, which have gained much attention to deal with such problems, are time-consuming. A challenging issue now is to propose solutions to this issue as affirmed by Prof Talbi (which

is one of the most renowned researchers in metaheuristics).

Also, I think that this new project http://demiurge.be/ proposed by IRIDIA tend to propose novel issues for both tuning and control of optimizers.

The best way to tune parameters of a metaheuristic algorithm is hit-and-trial. In other words, repeated runs with different values of parameters within specified range.

It depends on what you mean by parameter tuning, do you mean the algorithm itself being able to tune the parameters dynamically? Or do you mean how to set the parameters appropriately before hand?

Setting the parameters dynamically is mostly done based on the current performance of the algorithm (how successful it is in improving the population) along maybe with using some archiving methods. For a very successful and well cited algorithm, you can refer to LSHADE.

As for the setting the parameters beforehand, you would either have some mathematical model that can guide you to select parameter values that can guarantee convergence for example (a lot of work on PSO). Or you might have some heuristic rules-of-thumb to tune these parameters. If neither exists, then probably you need to employ some trial-and-error and try to identify the effects of the different parameters on the performance.

Salam,

@ Sir Nabab Alam, I think in this experimental design will be helpul. However, the purpose is to automate this process.

@Sir El. Abd, I think in most refereed studies, the term "tuning" corresponds to the first case and "control" to the second one as detailed in this book http://www.springer.com/us/book/9783642214332 (even if I have seen some seniors researchers in swarm intelligence consider the online setting as a tuning).

Best regards.

All the answers are great

Dear;

Please you can read it in the following link:

https://ieeexplore.ieee.org/document/7152377

Title of the article : Autonomous search and tracking of objects using model predictive control of unmanned aerial vehicle and gimbal: Hardware-in-the-loop simulation of payload and avionics.

Regards